4 Problems Converge on the AI Sector as Investors Await Earnings
Four issues have arrived at the same time. But they are not the same problem. Here’s what to know.
You've reached your free article limit
You've read 0 of 1 free Pro articles.
The semiconductor group is being hit hard Tuesday morning. South Korea’s Kospi fell more than 10% and triggered a temporary trading halt, Samsung Electronics dropped about 13%, and the benchmarks in Japan and Taiwan both lost around 4%. Nasdaq 100 futures are down about 0.75% while Dow futures are higher, which is continuation of the recent theme of technology underperformance.
This selling started in Asia and it is about Asian memory makers. The U.S. names are being dragged along rather than leading the way down, and so far the rest of the market is shrugging it off.
Four separate problems have arrived for the AI sector at the same time and they are getting blended together. They are not the same problem and have different ramifications for the stock market.
Competition Arrives in Memory
ChangXin Memory Technologies, China’s leading memory chipmaker, went public in Shanghai on Monday and jumped nearly 500% on its debut, briefly making it the most valuable company on that exchange. The market jumped to the conclusion that China will now be a serious competitor in memory chips and assigned a giant valuation.
That is a direct threat to the pricing power that has driven parabolic moves in the memory names. I have been discussing the bifurcation between the suppliers who capture the margin and the buyers who pay for it. The entire supplier side of that trade rests on scarcity but in a free market the issue of scarcity always attracts competition and that is how it ends.
Morgan Stanley warned Monday that contract chip prices should peak in late 2026. The Chinese listing is the market embracing the fact that competition is here now and pricing power is under attack.
The Financing Problem
Another issue being reported Tuesday morning is that the hyperscalers have borrowed more than $200 billion through bonds and loans this year with another $115 billion in announced equity raises. Surprise bond sales from Nvidia (NVDA) and Amazon (AMZN) pressured hyperscaler debt prices earlier this month.
The scale keeps growing. OpenAI is nearing a deal to lease a $500 billion data center in Ohio with $250 billion of financial backing from Nvidia. A supplier guaranteeing its customer’s financing creates an artificial boost to demand, and Bloomberg is now naming circular funding fears as a driver of this selloff. I discussed this on Monday as an issue that has surfaced before and faded but it is gaining traction again.
The Price of Money Is Rising
A third issue that is front-and-center into earnings is that Meta Platforms (META) priced a $12.5 billion offering for its Texas data-center project at a higher interest rate than a comparable deal last year.
Meta has real earnings, enormous cash flow, and a solid balance sheet but that apparently isn’t good enough. Lenders want more compensation to take the risk of funding Meta’s data centers. Everyone below Meta on the quality scale is going to pay even more.
This problem becomes even greater because the 10-year Treasury hit 4.7% last week, the highest since January 2025, and the spreads on hyperscaler debt have widened. A higher rate plus a wider spread is going to impact the ROI on all the CapEx spending.
Estimates May Be Engineered
A new issue is being reported this morning by The Wall Street Journal. There are questions about the numbers and estimates that are being used to justify the AI buildout.
Consensus views among analysts have the five big hyperscalers improving operating margins through 2029 to their best levels since Meta went public. Half of that improvement depends on overhead falling as a percentage of revenues but at the same time depreciation from all the new buildings and equipment will be significantly higher.
The argument is that the impact of this CapEx spending is being reverse engineered to fit the bullish thesis. Oracle (ORCL) is the example that is being used and we can see from the price action of the stock that there are concerns.
This concern is different from the other three. The first three worries are about the level of CapEx spending. This one is a growing concern that the estimates that are being used may be far too optimistic. Amazon, Alphabet (GOOGL) and Oracle are already free-cash-flow negative, and Alphabet has halted buybacks and issued equity to fund its spending. Chip prices may be coming down but the cost of infrastructure is going up.
What Settles It
We will have some hard evidence of the extent of these issues when Microsoft (MSFT) and Meta report Wednesday afternoon along with the Fed rate decision. SK Hynix (SKHY), Apple (AAPL) and Amazon follow Thursday, with Samsung reporting alongside them.
Despite all the consternation over AI estimates, the bright spot in all this is that S&P 500 companies are on track for earnings growth of about 38% year over year in the second quarter. That is far above what anyone expected coming in. The businesses are performing and the concerns are entirely about what the buildout costs and whether the returns arrive.
My Strategy
My posture has not changed and this action does not tempt me to change it. High cash, no significant technology exposure, and a shopping list of names outside the group that keeps developing better setups.
I have no interest in catching a falling chip stock in front of the Fed and four mega-cap reports. If Microsoft and Meta deliver and the reactions improve, there will be plenty of time to find entry points. If they disappoint into a market already worried about four separate things, the opportunities get better and the shopping list gets cheaper.
My plan is to stay vigilant, to watch chart development carefully and refine my shopping list as this substantial news flow is digested.
At the time of publication, Rev Shark had no positions in any securities mentioned.
